3.1 A corpus
In order to answer questions about grammatical changes which have taken place in the course of the twentieth century, I constructed a body or a CORPUS of data in the following manner. A type of data was needed which was likely to have remained fairly consistent throughout the century. The editorials or 'leading articles' in The Times of London seemed to fulfil this criterion. Accordingly, I chose a month at random (it happened to be March) and took the editorials for the first ten copies of The Times published in that month at five-yearly intervals. No Sunday papers were considered; they did not exist at the beginning of the time period. The years selected were 1900, 1905, 1910 and so on up until 1985 (1900 was, of course, not strictly speaking in the twentieth century, but this keeps the figures round ones). By this method, ten texts for each of 18 years were chosen, that is a total of 180 texts. Not all of these texts were of equal length, but even the shortest was made up of over two columns of newsprint. The amount of text considered, had it been printed in the same size of type as this book, would have covered over 500 pages.
It would be easy to find other ways of selecting appropriate data. The idea behind the system chosen
was that it might allow gradual developments to be seen, while at the same time allowing a fairly large body of text from each period to be considered. It will be seen in what follows that these goals were only partly achieved. To a certain extent, this can be attributed to Murphy's Law, which applies to corpus studies as to other aspects of life. When applying to corpora, it states that a corpus will never be the
right size for showing what you are trying to show: either it will be a bit too small, or it will be too big, and there will be too much data for easy analysis. The same corpus can be too small for some purposes and too large for others. This does not, however, mean that corpora are not useful in linguistic research; in many cases they are the only way of finding reliable data. What it does mean is that you have to evaluate the reliability and suitability of a particular corpus with regard to a particular point of interest when you consider the results that are obtained. The corpus taken from leading articles in The Times is no different from other corpora in this respect.
Other corpora of data will also be referred to in what follows, but the data from The Times will be referred to as The Times corpus.
Q How would you go about looking for change in English grammar? Would you need a corpus of data? Why (not)? What benefits arise from using your method? A You would need some kind of corpus of data, though it might not be data of the same kind as that mentioned here. Even a series of anecdotes about what people used to say (and you can't assume that such anecdotes are accurate) provides data. Using a corpus of the type I used means that the data can be checked and the experiments replicated by others. It leads to relatively precise and objectively verifiable statements about change.
3.2 Comparative and superlative marking
In English there are two ways of marking the comparative and superlative of adjectives. Generally speaking, monosyllabic adjectives (except ones like marked, prized which are created from participles) add the affix -er for the comparative and -est for the superlative: small, smaller, smallest. Adjectives with three or more syllables add the word more for the comparative, and the word most for the superlative: important, more important, most important. This leaves disyllabic adjectives unaccounted for. Some disyllabic adjectives take -er, -est while others take more, most. Some vary between the two usages. Thus we might agree that we
would probably say happier, happiest rather than more happy, most happy, (where these are genuine comparatives or superlatives: see below) and more senseless, most senseless rather than senselesser, senselessest, but we might not agree about whether we actually say commoner, commonest or more common, most common.
In dealing with such constructions, there are a few points that must be taken into account. Most is also used in a construction which is not a genuine superlative, in a sentence such as That is a most interesting remark, where most is equivalent to very. Care must be taken to keep the two constructions apart. There are also constructions with more and most which are structurally ambiguous between marking comparison and not marking comparison. Consider the following examples from The Times corpus. The candidate with most valid votes (10 March 1925) means 'with most votes which are valid rather than 'with votes which are most valid', and is not a genuine superlative. In other examples it is impossible to tell. The French Government is anxious to have more practical support (10 March 1950) is ambiguous between
support which is more practical' and 'a greater amount of support which is practical'. Again, care must be taken with such examples.
Where genuine comparatives and superlatives are concerned, both Barber (1964, p. 131) and Potter (1969, pp. 146–7) agree that, in Barber's words, ‘-er and -est are being replaced by forms with more and mosť. Both of them agree, moreover, as to why this change is taking place. As Potter says:
This change may be seen as another manifestation of the trend from synthesis to analysis, or from complex to simple forms, which has been going on for thousands of years in the history of our language from Indo-European to modern English.
In other words, as Barber phrases it, English is losing its inflections: a complex, inflected word like commoner is being replaced by a sequence of simple, uninflected words like more common. Neither of these scholars backs these assertions up with anything other than impressions. Both cite the adjective common as undergoing this change; they cite only about a dozen such adjectives. As Strang (1970, p. 58) pertinently remarks:
Barber thinks there is an increasing use of more, most, rather than -er, -est, in comparison, in keeping with a trend which again goes back at least four hundred years; he may be right, but we lack precise numerical information on the subject.
Although the terms 'synthesis' and 'analysis' will be useful again later and will be used in a more general discussion in section 6.4.2, for the time being we shall avoid them, and use instead the rather more perspicuous terms 'suffixation' for a form like commoner and periphrasis' for a form like more common. Discussion will thus be in terms of suffixed comparison and periphrastic comparison. Apart from this change from suffixation to periphrasis in the marking of comparison on disyllabic adjectives, both Barber and Potter also comment on the increasing use of periphrasis with monosyllabic adjectives (forms like most just), and the increased use of periphrasis in expressions like most wellknown rather than best-known. Neither of these points will be considered in the text, but there are some very brief comments in the Notes section at the end of the chapter.
At the end of the nineteenth century, Sweet (1891) reports that disyllabic adjectives which are stressed on the second syllable take the suffixed comparative, as do ‘many disyllabic adjectives with the stress in the first syllable' (1891, p. 326), but that adjectives ending in -ish, -s, -st, -ful and -ive take the periphrastic comparative; all participle forms (even when monosyllabic), also take the periphrastic comparative; and disyllables stressed on the second syllable are more likely to take the periphrastic form if they end in a 'heavy consonant-group', a consonantal cluster. It will be noted that Sweet leaves a lot of room for alternatives. A grammar written in the 1980s, Quirk et al. (1985) gives a rather different picture. According to this source the monosyllabic adjectives real, right, wrong and like always take the periphrastic form; otherwise monosyllabics ‘normally' take the suffixed form, but may take the periphrastic form especially in constructions such as more - than NP VP or the more - the more. Some disyllabic adjectives such as eager and proper only take the periphrastic form, but most take either, though they are more likely to take the suffixed form if they end in an unstressed vowel, syllabic /l/ or /ə/ (i.e. /ə/ followed by an 'r' in the spelling, which is always pronounced in some varieties of English, but is pronounced
only when the next sound is a vowel in others). The use of the periphrastic form is more likely with adjectives ending in -ly than with those ending in -y.
The following disyllabic adjectives can occur with [suffixed] forms (as well as periphrastic forms, which seem to be gaining ground): quiet, common, solid, cruel, wicked, polite, pleasant, handsome.(Quirk et al., 1985, p. 462)
(Both these descriptions are abbreviated and slightly simplified from their original sources, but give a good idea of the descriptions provided.)
These descriptions are very different, both in their general appearance and in their predictions. For Sweet, likelier is more likely than more likely; for Quirk et al., either might be found, with more likely perhaps slightly likelier! What is not clear is whether we are seeing a description of a change, or whether we are dealing with a different (and probably improved) description of the same set of facts.
The Times corpus does not give a very clear answer to questions like this. There are 17 disyllabic adjectives in that corpus which appear with both periphrastic and suffixed comparison: ample, bitter, common, complete, costly, deadly, empty, friendly, kindly, likely, obscure, remote, robust, severe, simple, sober, wealthy. There is some evidence of a tendency for the periphrastic comparison to be used later in the century than the suffixed comparison, but it is no more than a tendency: wealthy, for example, does not follow this general trend. In any case, the evidence is rather patchy, since there are more cases of comparison attested in the early years of the century than in later years. This means that in many cases suitable evidence for particular words is simply not provided by the data (see the data presented in Table 3.1). Moreover, there is no clear trend observable when all the adjectives are considered as a group. The ratio of suffixed to periphrastic comparatives for the 17 words listed above is higher in the period 1930–55 than in either the period 1900–25 or the period 1960–85. The evidence, therefore, is far from clear, with only the slightest evidence of a change towards the periphrastic comparative in the majority of those words which are attested with both forms.
In order to test whether this general trend could be
discovered in American English, and also in the hope of providing a more conclusive set of data, I carried out another experiment. This one involved reading The New York Times for January 1900 and January 1989. Starting with the papers of 1 January, I simply read, noting down every disyllabic adjective in the comparative or superlative, until I had collected 300 tokens from each source. This brought me to the paper of 12 January 1900, and to the end of section B in the paper of 3 January 1989. (The discrepancy is due to the increased size of the papers; both collections included a Sunday paper.) Inevitably, there will have been forms that I missed in my reading, though this should not matter a great deal. The random nature of the sample thus collected means
that some adjectives simply did not occur in one or other of the samples, sometimes for obvious reasons, but more often simply by accident. For example, I think that every occurrence of the adjective dainty was collected from the advertisements for the department stores' January sales in 1900, and was used to refer to women's clothing; in 1989 one simply did not advertise women's clothing as being dainty. On the other hand, the fact that remote failed to occur in the 1989 sample is presumably purely accidental.
I shall not present the data collected in this way in a table corresponding to Table 3.1 because it was just as inconclusive. Even increasing the size of the sample to 400 tokens from each year did not add to the relevant information, and made it look likely that a corpus of over 1000 tokens for each year would be necessary to draw clear conclusions. The general tendency seen in the corpus from The Times could also be vaguely perceived in the data fromthe New York Times, but no more than that. The proportion of suffixed comparatives to periphrastic comparatives remained relatively constant in the two years, which suggested that there is no clear-cut switch to periphrastic comparison.
Closer examination of both the data from The Times corpus and from the corpus from The New York Times suggests, however, that the question may have been wrongly posed by scholars such as Barber and Potter. There may be an alternative, and more useful, explanation of the changes which they observe. This different pattern is observable in both The New York Times corpus and the Times corpus. Data from The Times corpus will be considered first.
Most of the disyllabic adjectives which take suffixed comparison end in -y, and most of those which take periphrastic comparison do not. In Table 3.2 a list of those adjectives which do not fit this generalization is provided.
Three points need to be made about the data in Table 3.2. First, all the adjectives which end in -y and which are attested with periphrastic comparison after 1930 in fact end in -ly, where that -ly is an adjective-forming suffix. We can therefore generalize for the latter portion of the century that the suffix -ly demands periphrastic comparison, but that otherwise all words ending in -y take suffixed comparison.
Such a generalization does not hold before 1935. Secondly, it is striking how many of the forms with suffixed comparison end in syllabic /1/ spelled -le. For the latter part of the century, this is also a rule, with just one notable exception attested in 1975. Finally, note from Table 3.2 that all the words attested with suffixed comparison in breach of the primary generalization after 1950 are also attested with periphrastic comparison. It would be nice if there was clear evidence that the three relevant words in this class were moving from periphrastic to suffixed comparison, but there is no such clear data. Bitter is attested only once with the suffixed comparative: in 1955. Remote shows vacillation all through the century, but is possibly moving towards periphrastic comparison. Evidence on robust is too scarce to show anything, since it is only attested after 1960. What we appear to have, then, is a situation where, at the beginning of the century, there are no general rules about how to form the comparison of a disyllabic adjective. By the end of the century, the rules are becoming more fixed. Disyllabic adjectives which end in the suffix -ly take periphrastic comparison, other adjectives ending in -y and also those
ending in syllabic -le take suffixed comparison, all others take periphrastic comparison except for a few remnants which still vacillate between the old irregular form and the new regular form.
A similar pattern emerges from The New York Times data. If we consider what percentage of the words showing suffixed and periphrastic comparison in 1900 and 1989 end in -y, we find the results given in Table 3.3. What we appear to see here is suffixed comparison becoming more and more restricted to disyllabic adjectives ending in -y. And if we consider the percentage of words ending in -ly which show suffixed and periphrastic comparison, we find the results given in Table 3.4. Here we see periphrastic comparison becoming the preferred way of making comparison for adjectives ending in the suffix -ly. The figures in Table 3.3 are not statistically significant, but do seem to show a trend when taken in conjunction with the other results presented here, especially since comparable figures from The Times corpus are significant. The figures in Table 3.4 are statistically significant (p<0.04).
As another way of seeing how accurate the descriptive rules given above are for English in the 1980s, consider the
following. The Wellington Corpus of Written New Zealand English comprises approximately 1.1 million running words of text written or published in New Zealand in the years 1986–90. Since New Zealand English is being described here, there is no necessary reason for it to fit generalizations based on British and American English. However, of 564 cases of comparison of disyllabic adjectives (excluding those that are -ed or -ing forms of verbs), 92 per cent fit the generalizations suggested above, only 8 per cent do not (the adjectives which do not fit include narrow, quiet and shallow all attested exclusively with suffixed comparison).
When I saw a bumper sticker in 1989 which announced proudly that 'I've been to the G ... Hotel, the remotest hotel in mainland Britain' I found the suffixed comparison worthy of note; this is one of the remnants which still vacillates in general usage, though I must be (subconsciously) operating on the new rules for this word. Barber and Potter were, at an earlier stage in the century, struck by the use of common with an periphrastic comparative. This one has settled down into the new paradigms in the course of the century, and, indeed, is only used with periphrastic comparison in my corpus from The New York Times, even for 1900. Because there is still variability in the way the general rules are applied, it is possible to find both innovative and conservative forms which still sound a little unusual: Would you care to hear my own plan just in case? It's modester but would cause less upheaval.
(WHTM, p. 63). Garishest winebar for miles.
(AATD, p. 35). Charlie Braine was more clever than he was given credit for.
(ROSJ, p. 167). The change in the course of this century appears to have been only incidentally an increase in the use of periphrastic comparison. Rather, the change has been a regularization of a confused situation, so that it is becoming more predictable which form of comparison must be used:
Q Put the words from Table 3.2 into invented sentences. Which sound normal, and which do not? Do your class-mates agree? Do you agree with the tentative conclusion reached in this section or not?
A If you do not agree, see if you can formulate a better generalization. Do the words you disagree about have anything in common?